Notes
Ecology, 92(4), 2011, pp. 994–999
Ó 2011 by the Ecological Society of America
Differential responses of vertebrate and invertebrate herbivores
to traits of New Zealand subalpine shrubs
ANDREW J. TANENTZAP,1,6 WILLIAM G. LEE,2 JOHN S. DUGDALE,3 BRIAN P. PATRICK,4 MICHAEL FENNER,5
SUSAN WALKER,2 AND DAVID A. COOMES1
1
Department of Plant Sciences, University of Cambridge, Cambridge CB2 3EA United Kingdom
2
Landcare Research, Private Bag 1930, Dunedin, New Zealand
3
Landcare Research, Private Bag 6, Nelson, New Zealand
4
Central Stories, P.O. Box 308, Alexandra, New Zealand
5
School of Biological Sciences, University of Southampton, Southampton SO17 1BJ United Kingdom
Abstract. Plant traits are influenced by herbivore diet selection, but little is known about
how traits are affected by different types of herbivores. We related eight traits of 27 subalpine
shrub species in South Island, New Zealand, to damage of these shrubs by introduced red deer
(Cervus elaphus) and native invertebrate herbivores using phylogenetically explicit modeling.
Deer preferentially consumed species that grew quickly, were low in foliar tannins, or had high
leaf area per unit mass. However, these traits did not trade off against each other; rather, they
could be related to different multivariate defense strategies. Although the proportion of leaves
damaged by leaf-chewing invertebrates also increased with stem growth, invertebrates did not
damage the same fast growing species as those preferred by deer. Other traits may also be
important in determining herbivore preferences, as suggested by the high proportion of
variation in herbivory explained by phylogeny. Last, we found that the composition of
invertebrate herbivore communities was more similar among closely related shrubs, and
consequently, the range of invertebrate–plant associations may change if introduced deer shift
plant composition toward slow-growing species. Overall, our results demonstrate the
importance of herbivore type and coevolved interactions for the adaptive significance of
plant traits.
Key words: Cervus elaphus scoticus; coevolution; plant defenses; plant-herbivore interactions; red deer;
South Island, New Zealand; species introductions.
INTRODUCTION
Trait-mediated forage selection by herbivores plays an
important role in evolutionary and ecological structuring of plant communities (Ohgushi 2005). Plants employ
various traits to tolerate or resist tissue loss from
herbivores, including high growth rates, and physical
and chemical defenses (Herms and Mattson 1992), but
the relative investment of resources toward defenses vs.
growth will be determined by how plants incur damage
(Strauss and Agrawal 1999). Significant differences may
thus exist between which plant traits are associated with
different types of herbivory (Kotanen and Rosenthal
2000). For example, plants browsed by vertebrates may
possess traits that allow for rapid regrowth and
Manuscript received 30 April 2010; revised 23 September
2010; accepted 27 September 2010. Corresponding Editor:
P. M. Kotanen.
6
E-mail: [email protected]
994
‘‘tolerance’’ of herbivores rather than the chemical
defenses and ‘‘resistance’’ of herbivores that would be
associated with invertebrate damage (Kotanen and
Rosenthal 2000). However, studies rarely consider
whether plant traits are influenced by different types of
herbivores (e.g., vertebrate browser vs. invertebrate leaf
chewer), and instead focus on the relationships between
traits and a single herbivore. If different herbivores are
associated with different plant traits, as suggested by
recent studies demonstrating strong plant–herbivore
coevolution (Futuyma and Agrawal 2009), changes to
herbivore communities can alter the functioning of plant
communities.
We used phylogenetically explicit modeling (Hadfield
and Nakagawa 2010) to relate eight physical and
chemical traits of 27 subalpine shrub species in South
Island, New Zealand, to damage of these shrubs by
introduced red deer (Cervus elaphus scoticus) and native
April 2011
NOTES
invertebrate herbivores. Developments in statistical
methods allow us to test whether traits that have been
traditionally related to herbivory evolved repeatedly
across a lineage or arose from a single evolutionary
event (Agrawal 2007). Phylogeny should explain little
variation in traits that are ancestrally shared (i.e., when
trait values are approximately constant along a deeprooted branch of a phylogeny) compared with where
consistent character states (e.g., high or low trait values)
have arisen many times across a phylogeny. The
convergence of multiple traits toward a consistent
defense strategy provides important insight into how
the expression of functional traits is shaped by
herbivores (Agrawal 2007).
We asked (1) what plant traits influence herbivory by
introduced red deer and native leaf-chewing insects; (2)
do these two herbivore types prefer the same forage
species; and (3) are more closely related shrub genera
associated with more similar communities of invertebrate herbivores? Our approach also allowed us to ask
to what extent is covariation between plant traits and
herbivore damage explained by plant phylogeny? We
predicted that plant lineage would explain a large
amount of variation in damage by leaf-chewing invertebrates since these herbivores coevolved with the native
flora, so shrubs would have repeatedly adapted to
defend themselves against herbivores (Agrawal 2007).
We did not expect traits to be adapted to deer herbivory
since large mammalian herbivores were only introduced
to New Zealand in the 1800s and so did not coevolve
with the native flora. However, if the native vegetation
responds to deer in a similar way that it responded to the
avian herbivores with which plants coevolved (Forsyth
et al. 2010), we predicted traits would have repeatedly
evolved in a directional trend (either toward high or low
values) despite plants evolving independently of deer.
Under these conditions, we expected phylogeny to
explain a large amount of variation in trait values.
METHODS
Analysis of plant traits.—In March 1998, we harvested
and removed mature fully expanded leaves from three to
five separate plants of 27 decumbent, subalpine, shrub
species in South Island, New Zealand (45814 0 S, 167833 0
E; Appendix A). Fresh leaf area was measured with a LiCor 3000A portable leaf area meter (Li-Cor Biosciences,
Lincoln, Nebraska, USA), and divided by leaf dry mass
to calculate specific leaf area (SLA). Samples were air
dried in the field and then oven dried at 458C for 72
hours. Foliar N and P were determined colorimetrically
following a Kjeldahl digest (Blakemore et al. 1987),
while condensed tannins and non-tannin phenolics were
determined colorimetrically after extraction with 50%
acetone (Broadhurst and Jones 1978). We also counted
concentric growth rings at 50-cm height and divided
diameter by plant age to estimate annual rates of
995
diameter growth. Finally, we calculated wood density as
the ratio of oven-dry mass of stems to fresh volume.
None of the traits we measured, except for N and P,
covaried strongly (for all, r , 0.70; Appendix B).
Assessment of herbivory.—We classified shrubs as
either browsed or unbrowsed by deer within 345 10 3 10
m vegetation plots randomly located below treeline in
December 1980 and January 1981. Since browse damage
was not measured at the same time as trait data, and
could thus vary over time, we classified species into
categories rather than use a continuous variable. Plants
browsed in 10% of plots were classified as preferred,
while all others were considered non-preferred (Appendix A). Reassuringly, our browse categories are broadly
consistent with other palatability studies derived in New
Zealand, including analyses of forage usage/availability
and cafeteria trials (Tanentzap et al. 2009a, b, Forsyth et
al. 2010), demonstrating the robustness of our classifications to spatiotemporal variation.
We recorded the proportion of leaves damaged by
invertebrates for the three to five plants per species
sampled in 1998 (mean total number of leaves per plant
6 SE ¼ 51 6 1). Invertebrate defoliation is visually
distinct from deer browsing at the individual-plant scale
and none of the plants sampled for invertebrate damage
showed signs of deer damage. We also reviewed the
number of invertebrate leaf-chewing genera that were
associated with each shrub genus, using a national
database of invertebrate collections (Martin 2007),
supplemented with other data sources, to generate a list
of genera within Fiordland (Appendix C).
Statistical analyses of herbivore trait preferences.—We
related plant traits to the preference classes of deer and
the proportion of leaves damaged by invertebrates using
generalized linear mixed models estimated with the
MCMCglmm function in R version 2.9 (R Development
Core Team 2009; see Appendix D for details). We
assumed that each measure of herbivory (yi ) for each
species i could be predicted as
yi ¼ li
logitðyi Þ ¼ l þ Ti a þ ai þ e
ð1Þ
where l is the global intercept, T is a row vector of plant
traits for species i, a is a vector of estimated parameters,
a is a random intercept that varies among species, and e
is the residual error, which was fixed (Appendix D). For
deer preferences, yi represents the probability of species i
being browsed, i.e., Pr( y i ) ¼ 1. To account for
phylogenetic similarity among shrub species, we assumed a ; N(0, Ar2a ), where A acts as a covariance
matrix that is derived from published order and familylevel phylogenetic relationships (Appendix E). Traits
were standardized to a mean of 0 and standard deviation
of 1. Since invertebrate leaf damage was measured on
the same shrubs as plant traits, we used plant-level trait
996
NOTES
FIG. 1. Box plots of plant traits that were more strongly
supported than the null model to explain whether 27 subalpine
species are preferred by red deer. The heavy line in each box
denotes the median; the central box denotes the inter-quartile
range; whiskers indicate the 10th and 90th percentiles. SLA
stands for specific leaf area.
data rather than species-level arithmetic means for these
models. We used a Markov chain Monte Carlo sampler
to estimate the posterior probability distribution of
model parameters (Appendix D).
We used a backward model selection approach by first
fitting a model with all plant traits included (hereafter
‘‘full model’’), and then estimating 95% credible intervals
(CIs) for parameters associated with each trait. We
sequentially removed the trait with the largest CI
overlapping zero until all effects included in the model
had 95% CIs that did not overlap zero (hereafter ‘‘final
model’’). We compared the final model to the null model
using the deviance information criterion (DIC). Models
Ecology, Vol. 92, No. 4
with a lower DIC by 3 are better supported. We also
estimated the proportion of variance explained by plant
phylogeny (C ), after removing variation attributed to
fixed effects (Appendix D). By definition, C describes the
correlation between any two observations in the same
group (i.e., between two plants of the same species or
two species of the same genus), and is analogous to other
measures of phylogenetic heritability (Hadfield and
Nakagawa 2010). Values of C range from 0 (phylogenetically independent traits) to 1 (traits strongly covary
with phylogenetic relatedness).
Correlations between deer and invertebrate herbivory.—We fit a model relating deer preference to the
proportion of leaves damaged by invertebrates to test
whether deer-preferred shrub species were also preferred
by invertebrate herbivores. We used the same mixed
model approach as that used to relate deer preferences
to plant traits.
Community analyses of shrub–invertebrate associations.—We tested the correlation between shrub relatedness and invertebrate community similarity using a
Mantel test of distance matrices. Statistical significance
was estimated by creating 10 000 random permutations
of the two matrices, each time calculating the Pearson
correlation coefficient (r). The proportion of permutations where r is greater than the (observed) empirical
correlation (ro) tests the null hypothesis that there is no
relationship between the two matrices, i.e., permutations
lead to equal numbers of coefficients smaller and larger
than ro. We derived a distance matrix of shrub
relatedness by calculating the pairwise distances between
each genus in our hypothesized phylogenetic tree (see
Appendix E for detailed methods). For each pairwise
combination of shrub genera, we calculated a BrayCurtis dissimilarity index for the number of species in
each genus of invertebrates (vegan package in R;
Oksanen et al. 2010). Bray-Curtis values range from 0
(two shrub genera have identical herbivore communities) to 1 (two shrub genera do not share any
invertebrate genera).
RESULTS
Deer-preferred species had lower tannin concentrations, higher SLA, or higher diameter growth than
species that deer did not preferentially browse (DDIC of
final vs. null model ¼ 11.2; Figs. 1–2). The same traits
were found to affect deer preferences if we removed the
phylogenetic covariance matrix, indicating that patterns
of herbivore–trait associations persisted even if we
assumed that traits had not adapted across our
phylogeny (Appendix D).
The proportion of an individual shrub’s leaves
damaged by invertebrates increased with its diameter
growth, but the final model was only slightly better
supported than the null model (DDIC ¼ 4.2; Fig. 2). The
mean phylogenetic effect was similar between models of
April 2011
NOTES
997
FIG. 2. Estimated effects of traits of 27 subalpine shrub species on deer preference and invertebrate leaf damage 695% CI
(credible interval). C represents the strength of phylogenetic signal for deer and invertebrate leaf damage.
invertebrate damage and deer preference (0.87 and 0.89,
respectively), but the 95% CIs varied less for invertebrates (Fig. 2). There was no relationship between
whether a species was preferred by deer and damaged by
invertebrates (95% CI for effect of invertebrate damage:
0.23–5.28; DDIC ¼ 1.4). Only two deer-preferred
species were also preferred by invertebrates (10%
leaves damaged), with 14 of the 27 study species
undamaged by either deer or invertebrates (,10% of
leaves damaged; Appendix A).
More closely related shrub genera were associated
with more similar communities of invertebrate leaf
chewers (r ¼ 0. 45, P , 0. 001). Shrubs within the same
family or order had more similar herbivore communities
(mean Bray-Curtis dissimilarity index ¼ 0.68) than
shrubs that were less closely related (mean Bray-Curtis
FIG. 3. Comparisons of invertebrate herbivore communities (using Bray-Curtis dissimilarity index) between all pair-wise combinations of
13 shrub genera. Branch lengths estimated for
shrub genera using hypothesized phylogeny
(Appendix E), with lowest-shared taxonomic
ranks among genera increasing with phylogenetic
distance: family (0.0–0.33), order (0.33–0.50),
class (0.50–2.0), kingdom (2.0). There were only
seven unique distances between genera in our
hypothesized phylogeny (Appendix E). The
heavy line in each box denotes the median; the
central box denotes the inter-quartile range;
whiskers indicate the 10th and 90th percentiles.
¼ 0.89; Fig. 3). For example, Halocarpus and Podocarpus, both Podocarpaceae, shared more herbivores (BrayCurtis ¼ 0.62) than either did with Dracophyllum, an
angiosperm genus in the Ericaceae (0.94 and 0.91,
respectively).
DISCUSSION
Plant responses to herbivory have been predicted to
differ between vertebrates and invertebrates, with
vertebrate herbivory more likely to be associated with
resilience conferred by high plant growth (Kotanen and
Rosenthal 2000). However, our results suggest that
slow-growing shrubs are less preferred by all herbivores.
Fast-growing shrubs may have evolved to tolerate
herbivory, but different fast-growing shrubs are damaged by different herbivores. Other recent phylogenetic
998
NOTES
analyses have suggested that plants and herbivores
adapt to each other through a process of escalation in
the potency and diversity of plant defenses (Futuyma
and Agrawal 2009). Therefore, defense traits are
unlikely to be similar among closely related plants that
are associated with different herbivore communities
(‘‘pairwise evolution’’). Our findings support the alternative hypothesis of diffuse evolution, where ‘‘tolerance’’
against different herbivores persists across a lineage and
the evolution of these strategies is more influenced by
the collective impacts of herbivores rather than their
distinct identities (Futuyma and Agrawal 2009).
Deer preferentially ate shrubs with high growth rates
or few constitutive defenses. However, our findings do
not support the idea of simple bivariate tradeoffs
between growth and defenses expected under classic
plant-defense theories (e.g., Coley et al. 1985), because
there were no negative relationships between traits
representative of different anti-herbivore strategies,
e.g., growth and tannins. The absence of these relationships may arise because plants simultaneously employ
multiple defense traits, organized into defense syndromes that minimize costs (including to growth) while
maximizing defenses (Agrawal and Fishbein 2006).
Nonetheless, different defense syndromes can trade-off
against each other if they represent distinct adaptive
strategies (Agrawal and Fishbein 2006), and may
provide an explanation for why few shrubs in our study
both grew quickly and contained high levels of
constitutive defenses. Our findings were also consistent
with the protein competition model (derived from the
carbon–nutrient balance hypothesis; Bryant et al. 1983),
which predicts that carbon-based secondary compounds
accumulate within plants as growth declines in N- but
not P-limited environments (Wright et al. 2010). Foliar
N:P ratios are ,14 for 24 of our 27 study species,
suggesting that growth is N limited (Aerts and Chapin
2000), and leading plants with low foliar N to increase
investment in constitutive defenses (Appendices A and
B). The importance of resource availability in determining the relative investment by plants in defensive traits
also provides an explanation for how unrelated species
converge on a similar suite of traits to minimize
herbivore damage (Agrawal 2007).
Host plant phylogeny explained a large proportion of
variation in herbivore damage relative to traits commonly implicated in plant defenses, i.e., phenolics. In
our case, phylogeny may represent a robust indicator of
unmeasured foliar traits that are closely linked to
herbivore palatability (Pearse and Hipp 2009); e.g., high
concentrations of phytoecdysteroids that would be
avoided by invertebrates occur in the conifers in our
study (Singh et al. 1978). The strong phylogenetic signal
also suggests that these unmeasured traits act as defenses
against introduced deer because they were important in
deterring similar forms of herbivory by the native avian
Ecology, Vol. 92, No. 4
megafauna with which plants evolved (Forsyth et al.
2010). For example, aversion by deer to consuming
leaves with low SLA may overlap with predicted diet
choice by extinct moa to avoid plants with small and
inaccessible leaves (Bond et al. 2004). SLA integrates
leaf thickness and tissue density, both of which increase
fracture toughness and the resistance of tissue to tearing
by herbivores (Kitajima and Poorter 2010), but we could
not separate their effects on SLA since both were highly
correlated across species in our data set (r ¼ 0.93).
Additionally, we cannot exclude the possibility of
‘‘coincidental defenses,’’ whereby trait values have arisen
independently of herbivory but coincidentally confer
tolerance and resistance to herbivores (previously
termed ‘‘neutral resistance’’ by Edwards [1989]). Low
SLA, for example, can evolve in response to nutrientpoor soils (Cunningham et al. 1999), as a result of tradeoffs with other unmeasured traits (e.g., leaf survival;
Shipley et al. 2006), or to prevent oxidative stress and
maximize water-use efficiency (Bussotti 2008).
High levels of deer browsing may potentially affect
the close relationship between invertebrate communities
and their host plants. Although deer select forage based
on different traits than invertebrates, preferential
removal of fast-growing species with thin leaves at high
deer densities can shift community composition toward
slower-growing species with tough leaves (Tanentzap et
al. 2009b), some of which appear less preferred by
invertebrates, e.g., Podocarpaceae. While our study is
correlative, it suggests that introduced herbivores may
alter evolved trait-mediated interactions between shrubs
and native invertebrates, and this can lead to structural
changes in ecological communities (Ohgushi 2005).
ACKNOWLEDGMENTS
We thank Anthony Ives and Jarrod Hadfield for valuable
statistical advice and two anonymous reviewers for helpful
comments that improved our manuscript.
LITERATURE CITED
Aerts, R., and F. S. Chapin, III. 2000. The mineral nutrition of
wild plants revisited: a re-evaluation of processes and
patterns. Advances in Ecological Research 30:1–67.
Agrawal, A. A. 2007. Macroevolution of plant defense
strategies. Trends in Ecology and Evolution 22:103–109.
Agrawal, A. A., and M. Fishbein. 2006. Plant defense
syndromes. Ecology 87(Supplement):S132–S149.
Blakemore, L. C., B. K. Daly, and P. L. Searle. 1987. Methods
for chemical analysis of soils. New Zealand Soil Bureau
Scientific Report 80. Department of Scientific and Industrial
Research, Wellington, New Zealand.
Bond, W. J., W. G. Lee, and J. Craine. 2004. Plant structural
defenses against browsing birds: a legacy of New Zealand’s
extinct moas. Oikos 104:500–508.
Broadhurst, R. B., and W. T. Jones. 1978. Analysis of
condensed tannins using acidified vanillin. Journal of the
Science of Food and Agriculture 29:788–794.
Bryant, J. P., F. S. Chapin, III, and D. R. Klein. 1983. Carbon/
nutrient balance of boreal plants in relation to vertebrate
herbivory. Oikos 40:357–368.
April 2011
NOTES
Bussotti, F. 2008. Functional leaf traits, plant communities and
acclimation processes in relation to oxidative stress in trees: a
critical overview. Global Change Biology 14:2727–2739.
Coley, P. D., J. P. Bryant, and F. S. Chapin, III. 1985. Resource
availability and plant antiherbivore defense. Science 230:895–
899.
Cunningham, S. A., B. Summerhayes, and M. Westoby. 1999.
Evolutionary divergences in leaf structure and chemistry,
comparing rainfall and soil nutrient gradients. Ecological
Monographs 69:569–588.
Edwards, P. J. 1989. Insect herbivory and plant defence theory.
Pages 275–297 in P. J. Grubb and J. B. Whittaker, editors.
Towards a more exact ecology. Blackwell, Oxford, UK.
Forsyth, D. M., J. M. Wilmshurst, R. B. Allen, and D. A.
Coomes. 2010. Impacts of introduced deer and extinct moa
on New Zealand ecosystems. New Zealand Journal of
Ecology 34:48–65.
Futuyma, D. J., and A. A. Agrawal. 2009. Macroevolution and
the biological diversity of plants and herbivores. Proceedings
of the National Academy of Sciences USA 106:18054–18061.
Hadfield, J. D., and S. Nakagawa. 2010. General quantitative
genetic methods for comparative biology: phylogenies,
taxonomies and multi-trait models for continuous and
categorical characters. Journal of Evolutionary Biology
23:494–508.
Herms, D. A., and W. J. Mattson. 1992. The dilemma of plants:
to grow or defend. Quarterly Review of Biology 67:283–335.
Kitajima, K., and L. Poorter. 2010. Tissue-level leaf toughness,
but not lamina thickness, predicts sapling leaf lifespan and
shade tolerance of tropical tree species. New Phytologist
186:708–721.
Kotanen, P. M., and J. P. Rosenthal. 2000. Tolerating
herbivory: does the plant care if the herbivore has a
backbone? Evolutionary Ecology 14:537–549.
Martin, N. A. 2007. Plant-SyNZe. Crop and Food Research,
Auckland, New Zealand. hhttp://www.crop.cri.nz/home/
plant-synz/database/databasehome.phpi
999
Ohgushi, T. 2005. Indirect interaction webs: herbivore-induced
effects through trait change in plants. Annual Review of
Ecology, Evolution, and Systematics 36:81–105.
Oksanen, J., F. G. Blanchet, R. Kindt, P. Legendre, R. G.
O’Hara, G. L. Simpson, P. Solymos, M. H. H. Stevens, and
H. Wagner. 2010. vegan: community ecology package. R
package version 1.17-0. hhttp://CRAN.R-project.org/
package¼vegani
Pearse, I. S., and A. L. Hipp. 2009. Phylogenetic and trait
similarity to a native species predict herbivory on non-native
oaks. Proceedings of the National Academy of Sciences USA
106:18097–18102.
R Development Core team. 2009. R version 2.9. R Project for
Statistical Computing, Vienna, Austria. hwww.r-project.orgi
Shipley, B., M. J. Lechowicz, I. Wright, and P. B. Reich. 2006.
Fundamental trade-offs generating the worldwide leaf
economics spectrum. Ecology 87:535–541.
Singh, P., P. G. Fenemore, J. S. Dugdale, and G. B. Russell.
1978. The insecticidal activity of foliage from New Zealand
conifers. Biochemical Systematics and Ecology 6:103–106.
Strauss, S. Y., and A. A. Agrawal. 1999. The ecology and
evolution of plant tolerance to herbivory. Trends in Ecology
and Evolution 14:179–185.
Tanentzap, A. J., J. N. Bee, W. G. Lee, R. B. Lavers, J. A.
Mills, A. F. Mark, and D. A. Coomes. 2009a. The reliability
of palatability estimates obtained from rumen contents
analysis and a field-based index of diet selection. Journal of
Zoology 278:243–248.
Tanentzap, A. J., L. E. Burrows, W. G. Lee, G. Nugent, J. M.
Maxwell, and D. A. Coomes. 2009b. Landscape-level
vegetation recovery: progress after 39 years of invasive red
deer control in the Murchison Mountains, New Zealand.
Journal of Applied Ecology 46:1064–1072.
Wright, D. M., G. J. Jordan, W. G. Lee, R. P. Duncan, D. M.
Forsyth, and D. A. Coomes. 2010. Do leaves of plants on
phosphorus-impoverished soils contain high concentrations
of phenolic defence compounds? Functional Ecology 24:52–
61.
APPENDIX A
Plant traits and herbivore preference for 27 subalpine shrub species, South Island, New Zealand (Ecological Archives E092-082A1).
APPENDIX B
Correlation matrix for eight plant traits (Ecological Archives E092-082-A2).
APPENDIX C
Associations among 71 genera of leaf-chewing herbivores and 13 shrub genera in South Island, New Zealand (Ecological
Archives E092-082-A3).
APPENDIX D
Additional methods for model estimation (Ecological Archives E092-082-A4).
APPENDIX E
Hypothesized species phylogeny for 27 shrub species, South Island, New Zealand (Ecological Archives E092-082-A5).